User repository on GitHub
Talk to your Mac. Local whisper.cpp + one Jev (TypeSafe) call per command + macOS automation.
- Forks
- 22
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Voice
A hands-free macOS voice assistant: whisper.cpp transcribes locally in about 100 ms, then one roughly 250 ms Jev request selects a typed action plus arguments, and code executes it against the Mac.
- Stars
- 121
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- kevinbadi
- Last verified
- Sep 2026
About this project
Talk to your Mac. You speak, it opens apps, types, searches, scrolls, presses keys. Everything runs locally except one 250 ms call to Jev (TypeSafe's System One model), which turns the transcript into a typed action plus typed arguments in a single fan-out request. Jev never generates text; code produces candidate values and Jev selects. Code owns execution.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
Highlights
- Pipeline: mic, energy VAD, whisper.cpp on Metal (~100 ms), one Jev request (~250 ms), then macOS actions and spoken replies via say.
- Handles opening apps, site-specific search, typing, about 45 keyboard shortcuts, scroll-wheel events, volume and media keys, and compound commands.
- Multi-step tasks port the jev-ultrafast agent loop from the DOM to the macOS Accessibility tree, offering only operations with a valid target.
- Before input the executor re-reads the target's role, label, value, and enabled state, and hit-tests its centre so a covered control is never clicked.
- Everything runs locally except the Jev call; code produces candidate values and Jev selects rather than generating text.
Quickstart
bash
cp .env.example .env
./scripts/setup.sh
jev --text "open chrome and go to youtube" --dry-runWatch out
MIT-licensed; requires macOS on Apple Silicon with whisper-cpp, ffmpeg, uv, and a TypeSafe API key, plus Microphone, Accessibility, and Input Monitoring permissions.
Questions about this entry
Short answers drawn from the linked project or page.
Not really. Everything runs locally — energy VAD, then whisper.cpp on Metal in about 100 ms — except one roughly 250 ms Jev request that turns the transcript into a typed action plus typed arguments. Jev never generates text; code builds the candidate values and the model selects among them, and code owns execution.
Only from a small text model if you set TEXT_MODEL_API_KEY. Otherwise Jev selects a value from spans cut out of your goal — sentences, comma clauses, labelled values, numbers as written — so nothing is generated. Across the agent loop the model never emits selectors, coordinates, scripts, or shell.
Before every input the executor re-reads the target's role, label, value, and enabled state, re-resolves its geometry, and hit-tests its centre, so a covered control is not clicked. Only operations with a valid target on the current screen are even offered, and controls behind an open modal are withdrawn from the menu.
Runs are bounded: 40 actions, 80 Jev calls, and three actions in a row that change nothing stop the run as BLOCKED. A DONE is treated as the model's claim rather than proof, and a BLOCKED under 50% confidence executes the runner-up operation from the same request instead of stalling.
On macOS Apple Silicon, copy .env.example to .env and add TYPESAFE_API_KEY, then run ./scripts/setup.sh. It installs whisper-cpp and ffmpeg, downloads the model, remaps Caps Lock to F18 via hidutil and installs a `jev` launcher. Grant Microphone, Accessibility and Input Monitoring to your terminal.
`jev` is hands-free with a wake word (default Alfred or Jarvis), `--hold` uses Caps Lock to talk, `--always-on` treats every utterance as a command, and `--ptt` is push-to-talk in the terminal. `--text "..." --dry-run` tests routing with no microphone.
It runs the jev-ultrafast agent loop over the macOS Accessibility tree: an indexed element table, one Jev request, then an executor. Operations are CLICK, TYPE_TEXT, OPEN_APP, PRESS_KEY, SCROLL_UP, SCROLL_DOWN, WAIT, DONE and BLOCKED, and runs are bounded to 40 actions and 80 Jev calls.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Voice-controlled computer use on a Mac
X: A computer assistant that listens
X: Speech to live slides
X: Jev for real-time speech
More Jev resources like this
Multi-agent trading research with optional Jev screening of StockTwits and Reddit posts for company relevance before sentiment analysis runs.
Jev Tools & CLIs#community#python#trading
Community
TauricResearch/TradingAgentsPython
TauricResearch/TradingAgentsPythonBrowser Use's ultrafast agent: Jev picks an operation and an element from an indexed action space in one request, and a small LLM writes text only when the action is TYPE_TEXT.
Jev Tools & CLIs#community#python#browser
Community
browser-use/jev-ultrafastPython
browser-use/jev-ultrafastPythonA macOS computer-use loop that OCRs the screen, asks Jev for the next action with typed Choices, and clicks, at about $0.0002 per decision without sending screenshots to a frontier model.
Jev Tools & CLIs#community#python#agents
Community
awlevin/typesafe-computer-usePython
awlevin/typesafe-computer-usePythonA cross-platform memory layer for AI coding agents: conversations from Claude Code, Codex, DeepSeek Harness, OpenClaw, and OpenCode become Markdown memories indexed in Milvus, with optional Jev reranking through the TypeSafe API.
Jev Tools & CLIs#community#python#agents
Community
zilliztech/memsearchPython
zilliztech/memsearchPythonA desktop AI assistant with a Jev / System One inline plugin for classification, routing, verification, and scoring without chat-style output.
Jev Tools & CLIs#community#python#desktop
Community
szczyglis-dev/py-gptPython
szczyglis-dev/py-gptPythonA Go-based team workspace with optional Jev AI for agent auto-routing, task auto-fill, auto-assign, and automation condition nodes.
Jev Tools & CLIs#community#go#agents
Community
Paca-AI/pacaGo
Paca-AI/pacaGoFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.